Guide · AI assistants
Getting cited by AI assistants
ChatGPT, Claude, Perplexity, Gemini: these assistants return a written answer rather than a list of ten links. They compose that answer and cite a few sources within it. For a page, position is now measured by presence in the answer, where it used to be measured by rank. This guide explains how that choice plays out, what makes a page citable, and why this channel should be measured rather than assumed.
An answer is composed, a ranking is ordered
A search engine orders. It places your page at a rank, and that rank can be measured: second, eighth, fortieth. A conversational assistant works differently. It writes an answer and attaches a small number of sources to it. Those few sources are all the places that exist: the page is either among them or not, and that is what's to be won.
This difference changes what a promise can cover. A rank is earned by degrees; a citation is decided all at once. The citation remains the assistant's decision, just as the top spot remains the engine's: it alone grants it. What can be worked on, however, is entirely known: make the page readable, state clearly what it asserts, and record what happens next.
How assistants read you
An assistant reads your page as it is written. It reads it through existing indexes — Google's and Bing's among them — and through its own crawlers, which fetch the URLs themselves. Both paths have the same entry condition: the page must be accessible and understandable as it is served.
That's the good news here. The work is indistinguishable from classic SEO: what makes a page citable comes down to the same discipline you already know. The five conditions below are the ones you were already checking — you're simply checking them for one more reader.
01
A crawler can read the page
Assistants read through existing indexes and through their own crawlers. A page a crawler can read is a page that can be cited: readability comes first, the quality of the text is judged after that.
See it: Request the page the way a crawler does: the response must come back 200, the header and the page must leave indexing open, and robots.txt must allow the path.
Do it: Open access, then announce the page to the indexes. The announcement is what shortens the wait: an announced page gets picked up sooner.
02
The text is already in the served response
Text present in the HTML response is read on the first pass. Text produced afterwards by a script depends on a second pass: the timing of that pass and what it brings back are the crawler's business. What is already in the file is available immediately, to every reader, human and machine.
See it: Fetch the page as it is served, with scripts idle. Your title, your h1 and your main text should already be there, spelled out in full.
Do it: Serve the text within the response. Our own site is published in full HTML, served as-is: what you read is in the file, and the motion is handled in CSS.
03
The title and description state the subject
An assistant composes its answer from what it understood of the page. A title that states exactly what the page is about and a description that summarises it give the reader — machines included — enough to know in one line what the page covers, and how far it goes.
See it: Read the title and description on their own, with the page closed. The question it answers should be clear to you from those two lines alone; whatever you read there, a machine reads too.
Do it: One title per page, one description per page, both written for the actual subject matter. SEO Records fixes them at the moment the page is served.
04
Structured data describes the actual page
Structured data states plainly what the page contains: an article, a product listing, a question and its answer, an author, a date. It labels the text; it is the machine-readable tag for it. Its effect follows its accuracy: when it agrees with the page, it does exactly the job expected of it.
See it: Compare every declared value against what the page displays. A price, a date, a name: each must appear identically in the visible text.
Do it: Accurate, complete data drawn from your real data. SEO Autopilot writes descriptions, FAQs and internal links from your real data and nothing else, with every value sourced.
05
Every quantified claim carries its source and its date
An assistant's answer prefers what it can attribute. A dated, attributable claim is easier to reuse than a general one. That's the difference between a sentence you can quote as-is and a sentence you have to rephrase at your own risk.
See it: Take each figure on the page and look, right beside it, for where it comes from and when it dates from. A figure that carries its source and its date can be cited as-is.
Do it: Source it, date it, or drop it. A figure whose origin can be stated in one line returns more than it costs.
This channel is measured on your side
This is where this subject differs from all the others. Plenty of claims circulate about assistants and what they bring in. We stick to what your site measures for itself. Only one thing counts for your site: what can be read on your site.
The measurement is taken at the network edge, where requests arrive. We record which robot addresses read which pages, we certify those reads against the declared IP address ranges — that is what separates an announced robot from a visitor claiming to be one — and we then look at which pages generate visits. The result is a dated report, page by page.
That is what GEO SEO Records does. It exists for WordPress, for PrestaShop and for the network edge, and it places a dashboard in your admin panel: you read for yourself who came by, on which pages, and when. It also flags the pages concerned to the indexes, and it steps aside for SEO Records when both are installed.
Record what robots read on your site
GEO SEO Records measures this channel at the network edge and delivers the reading in your admin panel. It sells from €490 for one year, in a single payment. Before that, The Verifier reads your address free of charge, in about ten seconds, and gives you the state of the signals that determine your readability.